NVIDIA GeForce GTX 690 vs NVIDIA Tesla K20m Comparison

Head-to-Head Benchmarks

The two GPUs split their head-to-head benchmark wins evenly, but the margins tell very different stories. In Geekbench OpenCL, the NVIDIA GeForce GTX 690 posts 17,399 points against the Tesla K20m's 16,241, a 6.7% advantage. That is a solid, but not overwhelming, lead for the dual-GTX 690 in compute workloads that scale well with its dual-GK104 configuration. The Tesla K20m counters decisively in Geekbench Vulkan, scoring 21,936 versus 16,675, a 31.6% swing in favor of the compute-oriented Kepler card. This is a massive reversal, showing that the Tesla's architecture and driver stack handle Vulkan's low-level API far more effectively than the GeForce 600-series card.

Looking at the broader database averages, the Tesla K20m holds a 19,089 average benchmark score, placing it in the 64th percentile of all GPUs. The GTX 690 averages 17,037, which lands in the 60th percentile. The Tesla's nearest rivals in the database include the NVIDIA GeForce RTX 4050 Mobile (19,049, +0.2%), the AMD Radeon RX 6600 (19,036, +0.3%), and the NVIDIA Quadro K6000 (19,030, +0.3%), while it sits just 0.4% behind the GeForce GTX 780 (19,164). The GTX 690's closest comparators are the AMD Radeon HD 7970M (17,019, +0.1%), the AMD Radeon RX 7600 XT (17,083, -0.3%), the NVIDIA Tesla M4 (16,932, +0.6%), and the NVIDIA GeForce RTX 3070 (17,208, -1%). These delta percentages show that both cards are clustered tightly with their immediate peers, meaning the performance gap between the two is more meaningful than their positions relative to the wider GPU landscape.

The OpenCL result favors the GTX 690 by a narrow enough margin that real-world differences would depend heavily on workload optimization. The Vulkan result, however, is a categorical win for the Tesla K20m. A 31.6% lead is not a marginal difference; it indicates a fundamental advantage in how the GK110 silicon handles Vulkan command processing and memory management. The database shows exactly one win for each card, so the head-to-head record is a draw, but the magnitude of the Tesla's Vulkan victory outweighs the GTX 690's smaller OpenCL edge when weighing overall capability.

Where Each One Wins

The NVIDIA GeForce GTX 690 is the pick for OpenCL-based compute tasks. Its 17,399 score in that test beats the Tesla K20m by 6.7%, and its average benchmark score of 17,037 places it in the 60th percentile. The card's dual-GK104 design, with 1,536 shading units and a 256-bit memory bus, clearly responds well to OpenCL's execution model. The 2 GB GDDR5 frame buffer running at 6 Gbps effective provides 192.3 GB/s of bandwidth, which is sufficient for many OpenCL kernels that do not exceed the memory capacity. If a workload is OpenCL-bound and fits within 2 GB, the GTX 690 delivers a measurable advantage.

The NVIDIA Tesla K20m is the winner in Vulkan scenarios. Its 21,936 Vulkan score is 31.6% higher than the GTX 690's 16,675, and this is the standout result in the entire comparison. The Tesla's 5 GB GDDR5 memory on a 320-bit bus offers 208.0 GB/s of bandwidth, and its 2,496 shading units provide substantial compute throughput (3.524 TFLOPS FP32 versus 3.130 TFLOPS for the GTX 690). The larger memory capacity also makes the Tesla better suited for workloads that exceed 2 GB, such as larger datasets in scientific computing or machine learning inference. The Vulkan result suggests that the Tesla's driver and hardware combination extracts more performance from the modern API, making it the safer choice for any Vulkan-centric application.

For pure compute density, the Tesla K20m also holds the edge in raw specifications: 2,496 shading units, 208 TMUs, 40 ROPs, and a 561 mm² die with 7,080 million transistors. The GTX 690 counters with 1,536 shading units, 128 TMUs, and 32 ROPs on a 294 mm² die with 3,540 million transistors. The Tesla has more of everything except clock speed, where the GTX 690's 915 MHz base and 1019 MHz boost outpace the Tesla's unspecified base and boost clocks. In practice, the GTX 690 wins where clocks and dual-chip parallelism help, while the Tesla wins where shader count and memory capacity dominate.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA Tesla K20m has an average benchmark score of 19,089, which is higher than the NVIDIA GeForce GTX 690's 17,037. The Tesla also sits in the 64th percentile of all GPUs, compared to the GTX 690's 60th percentile.

Q: How large is the Vulkan performance gap?

A: The Tesla K20m scores 21,936 in Geekbench Vulkan, while the GTX 690 scores 16,675. That is a 31.6% difference in favor of the Tesla, making it the largest margin in any single benchmark between the two cards.

Q: Does the GTX 690 win any benchmark?

A: Yes, the GTX 690 wins Geekbench OpenCL with 17,399 points versus the Tesla K20m's 16,241, a 6.7% advantage. The head-to-head record is one win each.

Q: What are the memory specifications of each card?

A: The Tesla K20m has 5 GB of GDDR5 on a 320-bit bus with 208.0 GB/s bandwidth. The GTX 690 has 2 GB of GDDR5 on a 256-bit bus with 192.3 GB/s bandwidth. The Tesla's memory clock is 1300 MHz (5.2 Gbps effective), while the GTX 690's is 1502 MHz (6 Gbps effective).

Q: How do their transistor counts compare?

A: The Tesla K20m uses a GK110 chip with 7,080 million transistors on a 561 mm² die, giving a transistor density of 12.6M per mm². The GTX 690 uses a GK104 chip with 3,540 million transistors on a 294 mm² die, with a density of 12.0M per mm². Both are built on TSMC's 28 nm process.

Q: Which card has more shading units?

A: The Tesla K20m has 2,496 shading units, significantly more than the GTX 690's 1,536. The Tesla also has more TMUs (208 versus 128) and more ROPs (40 versus 32).

Specification Differences

The two cards differ in nearly every measured specification except for architecture generation, process node, and API support. Both use NVIDIA's Kepler architecture, are built on a 28 nm process at TSMC, and support DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. Beyond that, the differences are substantial.

The Tesla K20m uses the GK110 chip with 7,080 million transistors on a 561 mm² die, while the GTX 690 uses the GK104 chip with 3,540 million transistors on a 294 mm² die. Transistor density is 12.6M per mm² for the Tesla and 12.0M per mm² for the GTX 690. Clock speeds differ: the Tesla lists memory at 1300 MHz (5.2 Gbps effective) with no base or boost clock given, while the GTX 690 has a base clock of 915 MHz and a boost clock of 1019 MHz, with memory at 1502 MHz (6 Gbps effective).

Memory configuration is a major split. The Tesla offers 5 GB of GDDR5 on a 320-bit bus with 208.0 GB/s bandwidth. The GTX 690 offers 2 GB of GDDR5 on a 256-bit bus with 192.3 GB/s bandwidth. Compute resources also favor the Tesla: 2,496 shading units, 208 TMUs, and 40 ROPs versus 1,536 shading units, 128 TMUs, and 32 ROPs. Pixel rate is 36.71 GPixel/s for the Tesla and 32.61 GPixel/s for the GTX 690. Texture rate is 146.8 GTexel/s versus 130.4 GTexel/s. FP32 compute is 3.524 TFLOPS for the Tesla and 3.130 TFLOPS for the GTX 690.

Power and physical specs differ as well. The Tesla has a 225 W TDP with 1x 6-pin and 1x 8-pin power connectors and a suggested PSU of 550 W. The GTX 690 has a 300 W TDP with 2x 8-pin connectors and a suggested PSU of 700 W. The Tesla is 267 mm (10.5 inches) long, while the GTX 690 is 279 mm (11 inches) long and also has listed height (111 mm, 4.4 inches) and width (38 mm, 1.5 inches) dimensions. The Tesla has no display outputs, while the GTX 690 has 3x DVI and 1x mini-DisplayPort 1.2. The bus interface is PCIe 2.0 x16 for the Tesla and PCIe 3.0 x16 for the GTX 690. The Tesla launched in January 2013, the GTX 690 in May 2012.

Architecture Differences

Both cards share the Kepler architecture, but they are fundamentally different implementations. The Tesla K20m is built on the GK110 chip, a large compute-focused design with 7,080 million transistors on a 561 mm² die. This is a full-size compute silicon with 2,496 shading units, 208 TMUs, and 40 ROPs, engineered for throughput in professional workloads. The GTX 690 is built on the GK104 chip, a smaller, gaming-oriented design with 3,540 million transistors on a 294 mm² die. It packs 1,536 shading units, 128 TMUs, and 32 ROPs, but it is a dual-GPU card, meaning two GK104 chips work in tandem.

The Tesla K20m belongs to the Tesla Kepler (Kxx) generation and is the successor to Tesla Fermi, with Tesla Maxwell following it. The GTX 690 belongs to the GeForce 600 series, succeeding GeForce 500 and preceding GeForce 700. The Tesla has no display outputs, reflecting its role as a compute accelerator for servers and workstations. The GTX 690 includes display outputs (3x DVI, 1x mini-DisplayPort 1.2), marking it as a consumer graphics card.

Memory architecture differs significantly. The Tesla's 5 GB GDDR5 on a 320-bit bus provides 208.0 GB/s of bandwidth, which is higher than the GTX 690's 192.3 GB/s from 2 GB on a 256-bit bus. The Tesla's larger frame buffer and wider bus make it more suitable for data-intensive tasks, while the GTX 690's higher memory clock (1502 MHz versus 1300 MHz) partially compensates for its narrower bus. The Tesla also uses PCIe 2.0 x16, while the GTX 690 uses PCIe 3.0 x16, which gives the GTX 690 a potential advantage in data transfer speeds with compatible motherboards.

The Tesla's power delivery is more conservative at 225 W TDP with a 550 W suggested PSU, while the GTX 690 draws 300 W and requires a 700 W PSU. The Tesla uses a single 6-pin and single 8-pin connector, while the GTX 690 requires two 8-pin connectors. The Tesla's dual-slot cooler is 267 mm long, the GTX 690's dual-slot cooler is 279 mm long. The Tesla's transistor density is slightly higher (12.6M per mm² versus 12.0M per mm²), reflecting the GK110's more complex layout.

The Verdict

The data points to a clear split based on workload. For OpenCL compute tasks, the NVIDIA GeForce GTX 690 is the better performer, with a 6.7% lead in that benchmark. Its 17,399 OpenCL score, combined with 3.130 TFLOPS FP32 and a 915 MHz base clock, makes it a capable choice for applications that leverage OpenCL and fit within its 2 GB memory. The GTX 690 also has the advantage of display outputs, PCIe 3.0, and a lower launch MSRP of 999 USD, making it a more versatile card for mixed use.

For Vulkan workloads or any task that demands more than 2 GB of memory, the NVIDIA Tesla K20m is the clear winner. Its 31.6% Vulkan advantage is the single largest performance gap in this comparison, and its 5 GB GDDR5 frame buffer with 208.0 GB/s bandwidth provides headroom that the GTX 690 simply cannot match. The Tesla's 2,496 shading units and 3.524 TFLOPS FP32 also give it higher raw compute throughput. The Tesla's launch MSRP was 3,199 USD, reflecting its professional positioning.

The average benchmark score favors the Tesla (19,089 versus 17,037), and the Tesla sits higher in the percentile ranking (64th versus 60th). The Tesla's nearest rivals are all within 0.4% of its average score, showing it is competitively positioned among its peers. The GTX 690's nearest rivals are also close, with the RTX 3070 just 1% ahead and the RX 7600 XT 0.3% behind.

Choose the GTX 690 if OpenCL performance, display connectivity, and a lower launch price matter more than memory capacity and Vulkan speed. Choose the Tesla K20m if Vulkan performance, larger memory, or higher compute density are the priorities. For a pure compute accelerator in a server or workstation, the Tesla's specs and benchmark results make it the stronger option. For a desktop card that can also output video, the GTX 690 offers a more balanced package. The head-to-head record is one win each, but the magnitude of the Tesla's Vulkan victory gives it the edge in overall compute capability.